I first heard about the full moon effect while training in Boston as an emergency physician. The theory is that more patients with psychiatric problems come to the emergency room when the moon is full, compared to when it is waxing or waning. The relationship between mental health and the moon made its way into our language long ago. The word lunatic is derived from the Latin lunaticus, meaning “struck by the moon.” This full moon effect has even been the subject of academic study. At least five research teams have investigated the correlation between the phases of the moon and visits to the emergency room. None found any evidence of association, though try telling that to the ER staff on a busy night when a packed ambulance bay is illuminated by the gray cast of the full moon.
But there is a rhythm and flow that is fairly predictable in an emergency room. It’s usually quiet until the late morning. The busiest times are from about noon to ten p.m. As noted earlier, on Thanksgiving and Christmas Day very few patients come in. Black Friday is just the opposite; the ER overflows with patients. When I worked in inner-city emergency departments, summertime was stabbing season. There was more happening outside, more places where people gathered, more alcohol flowing. “If it is hot enough to barbecue,” a patient of mine once told me as I sutured the knife wounds on his chest, “it’s hot enough to stab someone.”
Influenza is predictable, too, and that’s part of its mystery. It hits us with clockwork regularity, but we don’t know why. It appears in the fall and winter and then, like a hibernating bear in reverse, disappears in the spring. Other infectious diseases have a seasonality too. Polio outbreaks were seen in the summer months. Measles would also rise and fall with the seasons. These diseases have been virtually eradicated by vaccines, but the flu remains a ritual, which can lull us into a false sense of security. That’s what happened at the onset of the 1918 pandemic.
In September 1918 the Journal of the American Medical Association noted that a new and more virulent form of influenza had broken out in several American cities and many army camps. Experts recommended calm. Outbreaks appeared to be following the usual pattern of influenza, and the first wave had already practically disappeared from the Allied troops. Since influenza was seasonal, it would be gone by the spring.
“It should not cause any greater importance to be attached to it,” the Journal said, nor “arouse any greater fear than would influenza without the new name.” The pattern of the 1918 pandemic was fairly typical of the flu. But the number of lives it claimed was not.
Why is there a flu season? Why is it unusual to get the flu in the summer? The spread of influenza has long been suspected to hinge on the climate. For example, some scientists have suggested that the way in which air flows on the outer edge of our atmosphere could play a role in the seasonal appearance of the flu, pushing more virus particles down into the air we breathe. Beyond these atmospheric changes, other researchers have focused on the role that humidity plays in the seasonality of the flu.
It should be noted that the seasonality of flu cases does not happen everywhere. In the tropics, there is no such thing as a flu season. There, influenza is generally present at low levels throughout the year, although in some places it spikes during the rainy season. Only in temperate regions do flu outbreaks rise and fall with the seasons. These regions are north and south of the tropics, extending as far as the Arctic and Antarctic Circles. Temperatures vary widely within Europe, Canada, and the United States, as well as much of Russia, North Africa, and the southern tip of Australia. The farther from the tropics, the bigger the change in climate between winter and summer, and the greater the seasonality of the flu virus.
There are several possible explanations for why there is an influenza season. One of the best-known has to do with how we crowd together: it’s called the “indoor contagion theory.” In the winter, the theory goes, people spend more time indoors, and there is a greater likelihood of close person-to-person contact. Coziness and proximity encourage transmission of the virus, and the number of influenza cases rises. This is most noticeable in schools and on college campuses, where young, social people live, work, and move around in tight quarters.
This theory turns up on many websites as an explanation for the seasonality of the flu, providing yet another reason (as if you needed one) to be cautious about believing what you read on the internet. It sounds like a neat explanation, but when you dig deeper you find many problems. For most adults in the West, the amount of time we spend indoors with others does not vary across the seasons. We go to work year-round and, other than a lunch break outside if it is warm enough, we don’t vary the amount of mingling we do. While students return to school in August or early September, they don’t start feeling achy and feverish until November, with the rest of the population. In the summer, we tend to use public transportation more than in the winter, and while doing so are more likely to be sneezed or coughed on. Yet there are very few cases of influenza seen in the summer. It’s confounding. “If close contact were all, one would think that London Transport would ensure an all-the-year epidemic,” wrote the British virologist Christopher Andrews, who was part of the team that first identified human influenza A in 1933. Cruise liners operate year-round, and despite the close contact of passengers, the pattern of influenza aboard these floating cities follows that on land.
The British astrophysicist Fred Hoyle theorized that the flu is connected to sunspots, which are magnetic flare-ups that discolor the surface of the sun. To be sure, Hoyle was a rather controversial theorist. He once suggested that viruses and bacteria did not evolve on Earth but rather arrived by comet, as microscopic alien hitchhikers. Hoyle dismissed the widely accepted Big Bang theory and believed instead in a steady-state universe that had always existed. So he was not out of character when he proposed that pandemic influenza had something to do with solar activity. In 1990 he published a letter in the prestigious scientific journal Nature, in which he pointed to a relationship between sunspot activity and influenza outbreaks. He speculated that the two might be linked. Hoyle noted that a recent flu epidemic in Britain aligned with one of the biggest sunspot flare-ups on record, and provided a graph showing the relationship between worldwide pandemics and sunspot activity. Each peak in the sunspot cycle was accompanied by an influenza pandemic. Hoyle reasoned that the intense electrical flares pulsing from the sun into our orbit could drive charged particles of the virus down from the upper atmosphere and into our noses.
At this point, you’re probably rolling your eyes. But let’s take a deep breath in the name of science. Increased solar activity does indeed have an effect here on the ground. If the sun’s activity could, in the words of NASA, “blow out transformers in power grids,” shouldn’t Hoyle’s suggestion at least be entertained? Since we can measure the cycles of increased solar activity, should we include these in an explanation of the seasonality of the flu? The big problem with the sunspot theory is that the definition of a pandemic is so subjective that it can be manipulated to fit any model or argument. It is therefore not surprising that Hoyle’s theory remains on the fringes. Instead, epidemiologists are less concerned with the sun’s spots as an explanation of flu’s seasonality, and more focused on the sun’s light—and the way it controls our levels of vitamin D.
The vitamin D theory has to do with a loss of our immune function over the winter months. In winter in the Northern Hemisphere, the sun is at a lower angle in the sky, resulting in fewer hours of daylight. This leads to less melatonin and less vitamin D production, which leads to immune suppression. That makes us more prone to illness, and more likely to catch the flu. In other words, flu epidemics may have something to do with the length of the day and our exposure to the sun’s light.
Vitamin D plays a key role in our health. While you can get some of it from your diet, most of your vitamin D comes from sunlight. After we manufacture a kind of cholesterol called 7-dehydrocholesterol, it is transported to the skin, where the sun’s ultraviolet rays transform it into vitamin D. Vitamin D primes our white blood cells for action against invading microbes. Some of these white cells, called macrophages and natural killer cells, release peptides and cytokines into the cells that are infected with the flu virus or with the bacteria that follow. Without vitamin D, these white cells, which are the backbone of our immune system, don’t work very well. Indeed, they might not work at all. And without our natural killer cells doing their natural killing, we become vulnerable to all kinds of viral and bacterial diseases.
What happens in places where there is significantly less sunlight in the winter? I grew up in London, where the gloomy winter sunrise might be as late as eight a.m. and sunset as early as four p.m. Trudging to and from school in the dark was not just depressing—it was also dangerous for my immune system. The British have lower vitamin D levels than those who live in sunnier climes. By some measures, twice as many people die over the dark winter than when the sun shines. The problem is especially severe for Britain’s senior citizens. Their long-sleeved clothing, while protective against the cold, limits their exposure to whatever sunlight there is. Low vitamin D is also much more common in African Americans than it is in pasty English people like me. In fact, they are more than seven times more likely to have low levels, because the melanin in their skin reduces the ability of sunlight to convert 7-dehydrocholesterol into vitamin D. We don’t know if this results in a greater incidence of influenza among African Americans, but their mortality from pneumonia and influenza is 10 percent higher than in the white population. This, too, supports the observation that vitamin D, and therefore the sun and our relationship to it, plays an important role in modifying our immune response.
Forty years ago in the Soviet Union, researchers hypothesized that Russians living in the far north were more susceptible to the influenza virus during the short days of winter than in the sun-drenched summertime. To test this they gave two groups of patients a flu vaccine that contained a weakened form of the virus. One group got the vaccine in the summer, when the days were very long, and one got it in the winter, when sunlight near the Arctic Circle is scarce. They found that the winter group was eight times more likely to develop a fever as a side effect of the vaccine. Less sunlight means less vitamin D. Which means a poorer immune system. Which means more side effects from the influenza vaccine.
Vitamin D makes for a well-functioning immune system, so could extra amounts of it in our diet actually prevent the flu entirely? Perhaps. In one experiment, a group of schoolchildren in Japan were given either vitamin D supplements or a placebo. The vitamin D group had significantly fewer cases of influenza. However, a similar study of healthy adults in New Zealand failed to show any decrease in the number of viral infections. And when older adults took extra doses of vitamin D it didn’t improve their immunological response to the influenza vaccine, which is disappointing (especially for Britain’s senior citizens). Clinicians faced with these kinds of conflicting studies can pool all the results together and then analyze the findings. This is called meta-analysis. One such analysis pooled the results of eleven studies of vitamin D. It showed that vitamin D was indeed effective at reducing the number of flu-like illnesses, but that it was no guarantee you could prevent them altogether. In other words, you can ingest heaps of vitamin D and still end up catching the flu.
The vitamin D theory suggests that the seasonality of influenza is due not to a feature of the virus, but to a feature of our immune systems. If we could somehow maintain our defenses against the virus year-round, we would not experience a winter uptick in flu. In contrast, some researchers believe that its seasonality can best be explained by features that have nothing to do with our immune system, or the sun itself. Instead, they blame the weather.
* * *
Jeff Shaman, an epidemiologist at Columbia University, uses computational modeling to predict the next influenza hot zone. He started out as a geophysicist, studied immunology, and later worked in climate and atmospheric science. His doctorate focused on modeling mosquito-borne disease transmission and its relationship to the weather. It was while investigating the transmission of the West Nile virus that he started to think about the far less exotic—but far more common—influenza virus, and how it is affected by humidity.
In 2007 a group from New York’s Mount Sinai School of Medicine had looked at the role of cold air and humidity in the transmission of the flu virus. They used guinea pigs as their guinea pigs. These animals are very susceptible to infection by the human influenza virus. After placing cages of infected guinea pigs next to cages of uninfected ones, the researchers blew air from the former to the latter, while varying both the temperature and the humidity. They found that when the temperature and humidity were both low, there was a high rate of transmission. The virus became less transmissible, however, as the humidity and temperature increased. In fact, once the temperature reached a balmy 86 degrees Fahrenheit (or 30 degrees Celsius) there was no transmission of the influenza virus at all. The uninfected animals remained happy and healthy.
This finding led Shaman to study the effect of humidity on influenza, and to build computer simulations that predict the location of the next influenza outbreak. Scientists from the CDC were also interested in the relationship between humidity and influenza transmission. In one experiment, they created a simulated coughing machine using “a metal bellows driven by a computer-controlled linear motor.” They armed their cough-bot with various sizes of flu-coated particles and aimed it at a mannequin whose mouth was replaced with a particle counter. They recorded the amount of virus that was transmitted, then changed the humidity of the room and repeated the experiment. In low-humidity environments, the virus particles remained infectious up to five times longer than in high-humidity ones. So by keeping the humidity high, it was theoretically possible to reduce the amount of influenza in the air. Practically speaking, though, using humidifiers to defeat influenza is not a viable plan. Although a segment of the population may install and use humidifiers, their overall use is very limited. Humidifiers are likely to be very low on the list of priorities for those with tight budgets. And humidifiers are less commonly used in indoor public spaces—where we do most of our coughing and sneezing at one another.
If humidity helps to explain why influenza is seasonal, and humidity is something we report in the weather forecast, could we predict future outbreaks based on the weather? There are real challenges with this because the spread of the flu depends on many factors, each of which changes in ways that are not predictable. In one academic paper on the subject, Shaman and a colleague wrote that “infectious disease dynamics are nonlinear and intrinsically chaotic.” No academic journal would allow its authors to write the straightforward version: an infectious disease refuses to follow simple rules.
But even though weather forecasting is also nonlinear, we have incorporated it into our daily lives. It, too, follows very complicated rules and has shifting variables, which is why accurately forecasting even next week’s weather is a very hard thing to get right. The components of a flu forecast are similar to those of a weather forecast. Instead of tracking cloud formation, we measure humidity. Instead of mapping how heat moves through the atmosphere, we look at how the flu moves through a population. Instead of radar and satellites, flu forecasters rely on the microbial equivalent: throat cultures and rapid flu tests provided by emergency departments and doctors’ offices. This gives forecasters some understanding of the flu in real time. Just like a TV weather forecaster with a live radar, flu forecasters report on conditions as real-world observations become available, and they constantly recalibrate.
Storm warnings contain two or three different paths that the storm could take, each with its own likelihood of occurring. This is known as ensemble forecasting. The ensemble is based on dozens or hundreds of data points; each may predict slightly different outcomes, but when combined they produce a most likely and a least likely scenario, together with a few scenarios somewhere in the middle. Flu forecasters now produce ensemble forecasts of what could happen over the flu season. The final prediction consists of a few possible scenarios and how likely they are to occur.
Shaman and his colleagues used their forecasting model during the flu season that began in the fall of 2012. They estimated the spread of flu in 108 cities in the United States and produced weekly real-time forecasts. At first, the forecasts didn’t seem to be useful. Their overall accuracy was so low that if they’d been given by a weather forecaster, you would have turned to another channel. But the model rapidly improved as they added more data from the field. By the end of the season their weather model was about 75 percent accurate in its reflection of the flu. That’s not perfect, but it’s far better than predictions based solely on historical data.
Shaman’s flu forecasting success that year caught the eye of the CDC in Atlanta. In 2014 they declared him the winner of their Predict the Influenza Season Challenge—an accolade that came with a prize of $75,000. Based on his successes, Shaman has big dreams for the future of flu forecasting. During the flu season, he would like to see a nightly flu forecast alongside the familiar weather forecast we already get on the ten o’clock news. This isn’t as odd as it sounds. After all, pollen counts and pollution warnings have long been part of our TV weather reports.
As an ER doctor, I’m not sure what I could do with a flu forecast. If the weather forecast predicted an 80 percent chance of rain, I would leave home carrying an umbrella. But what on earth would I do with a flu forecast that predicted with 80 percent certainty that the flu season would peak in a week?
With an accurate flu forecast, Shaman hopes that hospitals would change their staffing patterns and, if things look really bad, prepare extra equipment like ventilators. In a normal flu season, hospitals can certainly cope with the few additional patients who have respiratory failure due to a complication of influenza. But in a pandemic on the scale of the 1918 outbreak, an enormous number of patients would need help.
As an example, take metropolitan Atlanta and a hypothetical flu pandemic lasting eight weeks. At the peak of the outbreak, an estimated 2,000 patients a week would require admission to the hospital and more than three-quarters of the beds in the city’s intensive care units would be occupied by patients with influenza. Almost half of the existing ventilators would be dedicated to keeping the sickest of them alive—that’s in addition to the ventilators needed for patients who are in the ICU for other reasons. This is where Shaman’s flu forecasting might be very useful. Because it estimates the number of likely cases and when the pandemic will peak, it would give hospital administrators and public health officials time to plan ahead.
It’s a great idea, but I doubt it would work, because of the realities of hospital management. I’ve been in emergency departments for years but have yet to see any of them change their staffing patterns or medication supplies in response to a bad flu outbreak. Hospitals rarely take these steps, which are either costly, impractical, or both. Whose surgery should they cancel to make a bed available in the intensive care unit for a flu patient who might never arrive? Many emergency departments are already operating at maximum capacity and hospitals are short of nurses. There is not much wiggle room to make more beds available. A hospital bed is like a plane seat: it generates no revenue if it is not filled. That drives hospitals to be at or near 100 percent bed capacity. Asking a hospital to keep beds open and staffed for a flu epidemic that hasn’t yet arrived is like asking an airline to keep ten rows of seats unoccupied for standby passengers who may never show.
When Shaman discussed his flu forecast with public health officials, they were skeptical. They suggested that he forget the whole idea and instead just encourage people to get vaccinated. But Shaman believes his flu forecast can target vulnerable populations and help increase the rate of vaccination in the United States, which is around 40 percent in adults. Once you receive the vaccine, your immune system takes a couple of weeks to mount an adequate response. Timing is essential. Public health campaigns that encourage immunization might be improved if they were based on the actual risk of influenza for that year.
Here, too, there are lessons to be learned from one kind of rather severe weather forecast: hurricane prediction. Communities exposed to warnings that incorrectly predict a hurricane change how they respond to these warnings in the future. They are less likely to listen to them. Public health advisories about the flu would be more effective if they were coupled with an accurate forecast of how bad the season will be. The message would not simply be “Get vaccinated.” Instead it would be “Get vaccinated now in time for the peak in flu cases, which is predicted to occur in ten days.”
The seasonality of the flu is less of a mystery than it once was thanks to the work of Jeff Shaman and others. Humidity, sunlight, and temperature each seem to play a role, but as anyone working in the field will tell you, there is more waiting to be discovered. Our ability to accurately forecast the flu’s peaks and troughs seems to be improving, but a daily flu forecast is not around the corner. Perhaps the best way to battle influenza is to forget about when it will make its move and head it off at the pass, to prevent its spread in the first place. Drugs exist to do just that, and they are so important that they’re kept in secret stockpiles. They are prized and protected—and enormously profitable for their makers—but do they revolutionize our fight against the flu, or merely act as a security blanket that reassures rather than cures?